Un criterio generalizzato per l'imputazione di dati mancanti in indagini congiunturali

Authors

  • Roberto Gismondi ISTAT, Istituto Nazionale di Statistica

DOI:

https://doi.org/10.6092/issn.1973-2201/1105

Abstract

The analysis of non response effects in longitudinal and conjunctural surveys is a widely discussed topic. Neverthless, use of traditional tools originally thought of for cross section frameworks could cause bad performances and often ignores additional information concerning units under observation. In this paper, after a short resume about advantages and risks due to the use of imputation techniques or subjective matching rules for estimating missing values, we introduce a simple estimation criterion based on the well known donor method. Original contributions concern the attempt to generalize a previous version of the proposed estimator; the possibility to take into account both the previous response pattern for non respondents and their features, in terms of some known quantitative variables picked up from the same survey or coming from other sources; a wide discussion on practical problems that occur while attempting to calculate the estimates for missing values; finally, proposal on qualitative indexes to be computed to better analyze the global efficiency of the procedure. The article ends with a concrete application on hypermarkets monthly sales for the year 1996, from which other useful considerations have been drawn in order to steer additional theoretical proposal.

How to Cite

Gismondi, R. (1999). Un criterio generalizzato per l’imputazione di dati mancanti in indagini congiunturali. Statistica, 59(1), 83–100. https://doi.org/10.6092/issn.1973-2201/1105

Issue

Section

Articles